Shulou(Shulou.com)06/01 Report--

Using fragmented time to view information has become a way of life for Internet aborigines. According to big data's service QuestMobile, the size of the integrated information industry has reached 530 million at the end of 2016, a year-on-year growth rate of nearly 70 percent. More than half of mobile netizens are using News App.

QuestMobile released its Spring 2017 report on April 18. Sina News App, one of the most important branches of Sina, has 60.56 million monthly users (monthly active users), an increase of 138.3% over the same period last year, with medium and high consumption level accounting for 75.4%. In the face of such a mass, how does the back-end database of Sina support? What are the criteria for database selection? How to control data security? With these questions, we visited Zhao Jingbo, senior DBA of Sina database platform.

Zhao Jingbo has successively worked in Hangzhou Walk Technology and Sina, engaged in the operation and maintenance of Oracle, MySQL and Redis. At present, it is the senior DBA of Sina database platform, which is mainly responsible for the automatic operation and maintenance of Sina database platform Redis, kafka and other related aspects.

Reveal the composition of the database of Sina

According to Zhao Jingbo, at present, Sina's database platform has 9 major IDC,1200+ servers, 7k + instances, 100 billion hits/ days, and a total storage capacity of 1PB +. In order to better meet the company's business needs, Sina has adopted different database platforms and data system service platforms to provide related data services:

Storage Service of Relational Database based on MySQL

KV Storage Service based on Redis

Document Storage Service based on Mongodb

Semi-structured big data storage service based on Hbase

Caching Service based on Memcached

Message queuing Service based on mcq and kafaka

Zookeeper service, you can also apply for data publication / subscription, naming service, election and other needs.

As a mainstream news portal in China, Sina has a wide range of business scenarios. How do different business scenarios select the database? Zhao Jingbo said that because MySQL is a very mature product, its plug-in engine features can also meet the requirements of most scenarios, coupled with its active community and the most abundant talent reserve, MySQL will be preferred if the business scenario is appropriate. However, if the business scenario has some special requirements, we will also consider other databases, such as the business scenario is only a simple kv storage model, or the business requires very high response time, or the business needs a database of schemaless type to facilitate the rapid development and iteration of the business, and so on.

At present, Sina platform not only provides MySQL services, but also provides MongoDB, Redis, Memcached, Hbase and other services. Zhao Jingbo introduced the changes of the internal service publishing system of Sina. In the early years, all the news released by Sina was stored in MySQL. With the development of the business, the personalized demand is increasing day by day, and the demand for business table structure change (adding fields) is also more and more. At this time, the storage model of MySQL fixed schema can not meet the needs of rapid business development iteration, so all the news data of the publishing system is migrated from MySQL to MongoDB. In addition, news recommendations such as push require high response time for some interfaces, so the in-memory KV cache such as Redis or Memcached is used to solve the business requirements.

How to keep the control data safe on Sina

For Internet companies, data security can be said to be a lifeline. According to relevant media reports, in the first 10 months of 2016 alone, there have been about 3000 public data leaks worldwide and 2.2 billion records have been disclosed, more than the whole of 2015. In the face of such a severe situation of data security, Zhao Jingbo said that Sina has many rigid regulations on data security, and has made a hierarchical division of host dimension, business dimension and service dimension.

Host dimension: the relevant standards are established, and the baseline requirements of host security protection are given, including authentication, access control, intrusion prevention, etc., as well as strict standards for installed software, login methods, and so on.

Business dimension: it defines the security protection requirements of Sina business and applications, which mainly restricts the uploads and downloads of applications, links outside the website, malicious code, information encryption and so on.

Data service dimension: the security at this level is mainly reflected in database backup, storage reliability, server resource use (isolated through Cgroup), database access account, SQL injection prevention, data encryption, data consistency and so on.

The dynamic password update mechanism is established to update the database account password dynamically for the very core key business, so as to improve the security.

The database account only retains the rights to add, delete, modify and check, and other permissions such as truncate and drop are all reclaimed, which greatly reduces the chances of the business side mistakenly deleting data through such dangerous commands.

To classify the accounts of the internal management database, only a few senior and senior DBA have greater permissions to operate the database, and conventional DBA only retains binding permissions on fixed servers.

Data storage security mainly does two things. One is for confidential data, and we encrypt the backed up data. Except for the corresponding full-time administrator, others do not have the permission to view it. The other is that the data we backed up will be stored in the HDFS cluster to ensure the reliability of the data.

In the early stage of business development, in order to meet the requirements of business resource applications more quickly, ignoring the normative construction of the platform itself is something that many enterprises are facing or are experiencing. At the China Database Technology Conference (DTCC 2017), Zhao Jingbo will take Redis service as an example to tell us what pain points Sina's core business operation and maintenance have encountered with the growth of business scale. When more and more problems are exposed in the process of operation and maintenance, how to standardize and improve the Redis operation and maintenance platform.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.